Research map / 3D vision / Human pose estimation
TAIHRI: Task-Aware 3D Human Keypoints Localization for Close-Range Human-Robot Interaction
Ao Li, Yonggen Ling, Yiyang Lin, Yuji Wang, Yong Deng, Yansong Tang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add camera intrinsics unification (fixed focal length) (ablation) | TAIHRI · Harmony4D-Egocentric (Upper) | G-MPJPE (mm): 425.13 -> 93.83 | Improved |
| add camera intrinsics unification (fixed focal length) (ablation) | TAIHRI · Harmony4D-Egocentric (Lower) | G-MPJPE (mm): 433.59 -> 114.98 | Improved |
| use focal-length unification instead of learnable ray embedding (ablation) | TAIHRI · Harmony4D-Egocentric (Upper) | G-MPJPE (mm): 380.29 -> 93.83 | Improved |
| use focal-length unification instead of learnable ray embedding (ablation) | TAIHRI · Harmony4D-Egocentric (L-Upper) | G-MPJPE (mm): 400.09 -> 107.81 | Improved |
| add 2D keypoint reasoning before 3D prediction (ablation) | TAIHRI · Harmony4D-Egocentric (Upper) | G-MPJPE (mm): 126.67 -> 93.83 | Improved |
| add 2D keypoint reasoning before 3D prediction (ablation) | TAIHRI · Harmony4D-Egocentric (Lower) | G-MPJPE (mm): 138.39 -> 114.98 | Improved |
| add reinforcement finetuning (GRPO) after SFT (ablation) | TAIHRI · Harmony4D-Egocentric (Upper) | G-MPJPE (mm): 101.82 -> 93.83 | Improved |
| add reinforcement finetuning (GRPO) after SFT (ablation) | TAIHRI · Harmony4D-Egocentric (R-Upper) | G-MPJPE (mm): 110.79 -> 98.23 | Improved |
| use Huber+PCK reward instead of MSE reward in RFT (ablation) | TAIHRI · Harmony4D-Egocentric (Upper) | G-MPJPE (mm): 795.24 -> 93.83 | Improved |
| scale backbone from Qwen3-VL-2B to 4B | TAIHRI · Harmony4D-Egocentric (L-Upper) | G-MPJPE (mm): 119.86 -> 107.81 | Improved |
| scale backbone from Qwen3-VL-2B to 4B | TAIHRI · EgoBody test sample (Upper) | G-MPJPE (mm): 85.62 -> 75.77 | Improved |
| use TAIHRI instead of SAM 3D Body (3DB-H) | close-range keypoint localization · Harmony4D-Egocentric (Upper) | G-MPJPE (mm): 127.72 -> 93.83 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PromptHMR — PromptHMR: Promptable Human Mesh Recovery (CVPR 2025) · better on 5 of 5 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- InteractVLM: 3D Interaction Reasoning from 2D Foundational Models (CVPR 2025 · similarity 0.87)
- HandVQA: Diagnosing and Improving Fine-Grained Spatial Reasoning about Hands in Vision-Language Models (CVPR 2026 · similarity 0.85)
- InteractAnything: Zero-shot Human Object Interaction Synthesis via LLM Feedback and Object Affordance Parsing (CVPR 2025 · similarity 0.84)
- Neural Localizer Fields for Continuous 3D Human Pose and Shape Estimation (NeurIPS 2024 · similarity 0.84)
- Forecasting of 3D Whole-body Human Poses with Grasping Objects (CVPR 2024 · similarity 0.84)
- OpenHOI: Open-World Hand-Object Interaction Synthesis with Multimodal Large Language Model (NeurIPS 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
